What is the core idea behind generative AI?
Generative AI does not retrieve answers, it produces new content by predicting what is most likely to come next.
How does generative AI differ from related concepts?
| Concept | Difference |
|---|---|
| Generative vs Discriminative AI | Discriminative models classify or label existing data. Generative models create new data. |
| Generative AI vs LLMs | Large language models are generative AI specialised in text. Generative AI also covers images, audio, and video. |
| Generative AI vs Traditional AI | Traditional AI follows rules or predicts categories. Generative AI synthesises original output. |
How does generative AI work?
- A model is trained on large datasets to learn statistical patterns
- That knowledge is stored in billions of internal parameters
- Given a prompt, the model predicts and assembles new content step by step
- Techniques include transformers for text and diffusion models for images
What can generative AI create?
- Text, from articles to code, through large language models
- Images and video, through diffusion models
- Audio, speech, and music
- Synthetic data used to train other models
Why is generative AI important?
Generative AI is the technology behind ChatGPT, Claude, Gemini, and image tools like Midjourney and DALL-E. It moved AI from analysing data to producing work, reshaping software, media, and knowledge work across the economy.
How is generative AI used in practice?
Common uses include drafting and editing text, writing and reviewing code, generating marketing and design assets, summarising documents, and powering conversational assistants. Its main limitations are hallucination, bias inherited from training data, and cost at scale.